3D variational brain tumor segmentation on a clustered feature set

Abstract

Tumor segmentation from MRI data is a particularly challenging and time consuming task. Tumors have a large diversity in shape and appearance with intensities overlapping the normal brain tissues. In addition, an expanding tumor can also deflect and deform nearby tissue. Our work addresses these last two difficult problems. We exploit the various MRI modalities and their texture characteristics to construct a multi-dimensional feature set. Further, we extract clusters which provide a compact representation of the essential information in these features. The main idea in this paper is to incorporate these clustered features into the 3D variational segmentation framework. In contrast to the previous variational approaches, we propose a segmentation method that evolves the contour in a supervised fashion. The segmentation boundary is driven by the learned inside and outside region voxel probabilities in the cluster space. We incorporate prior knowledge about the normal brain tissue appearance, during the estimation of these region statistics. In particular, we use a Dirichlet prior that discourages the clusters in the ventricles to be in the tumor and hence better disambiguate the tumor from brain tissue. We show the performance of our method on real MRI scans. The experimental dataset, includes difficult instances with tumors that are inhomogeneous in appearance, small in size and in proximity to the major structures in the brain. Our method shows good results on these test cases.

A Survey of MRI Segmentation Techniques for Brain Tumor Studies

“…This method requires the image to be segmented to contain distinct features that are quantifiable that allows correct labeling of image pixels by the training data. The clustering algorithm is trained to learn on its own from the available data to deduce the segmentation [38]. Clustering methods do require an initial segmentation to identify the different classes in an image and they are highly sensitive to this initial segmentation.…”

“…Clustering methods are sensitive to noise and intensity in homogeneities as they do not model spatial interactions. But absence of spatial modeling has helped in producing faster computation results [38].…”

Brain Tumors: How Can Images and Segmentation Techniques Help?

“…Finally, the geometric level set method was applied with the same evolution strategy used in (Ho et al, 2002) and also prior knowledge was included to refine the segmentation of tumor and edema. Similar approaches were proposed by (Cobzas et al, 2007) and (Popuri et al, 2009), but instead of using a level set for refinement in a final step as done by (Prastawa et al, 2004), a feature set of specific anatomical priors was fully integrated into the region-based variational formulation. (Xie et al, 2005) presented a hybrid approach, i.e., the evolution of the curve is governed simultaneously by region and boundary information that serves as propagation force and stopping function, respectively.…”

A Neural Network based Method for Brain Abnormality Detection in MR Images Using Gabor Wavelets

“…The analyzing methods have been done until now used values of pixels intensities, pixels coordinates [8,18,19] and some other statistic features such as mean, variance or median which have much error in determination process and low accuracy and robustness in classification [8][9][10][11][12][13][14][15][16][17][18][19].…”